BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:AcousNomaly: Learning to Detect Anomalies in Acoustic Telemetry Da
 ta using Machine learning and Deep Learning
DTSTART;VALUE=DATE-TIME:20241204T133000Z
DTEND;VALUE=DATE-TIME:20241204T135000Z
DTSTAMP;VALUE=DATE-TIME:20260906T111835Z
UID:indico-contribution-2419@events.chpc.ac.za
DESCRIPTION:Speakers: Siphendulwe Zaza* (MSc Applied Mathematics student f
 rom Rhodes University)\nAcoustic telemetry data plays a vital role in unde
 rstanding the be-\nhaviour and movement of aquatic animals. However\, thes
 e datasets\,\nwhich can often consist of millions of individual data point
 s\, often\ncontain anomalous detections that can pose challenges in data a
 nalysis\nand interpretation. Anomalies in acoustic telemetry data can occu
 r due\nto various biological and environmental factors\, and technological
  limi-\ntations. Anomalous movements are generally identified manually\, w
 hich\ncan be extremely time-consuming in large datasets. As such\, this st
 udy\nfocuses on automating the process of anomaly detection in telemetry\n
 datasets using machine learning (ML) and artificial intelligence (AI)\nmod
 els. Fifty dusky kob *(Argyrosomus japonicus)* were surgically fit-\nted w
 ith unique coded acoustic transmitters in the Breede Estuary\,\nSouth Afri
 ca\, and their movements were monitored using an array of\n16 acoustic rec
 eivers deployed throughout the estuary between 2016\nand 2021\, resulting 
 in more than 3 million individual data points. The\nresearch approach comb
 ined the use of Neural Network (NN) models\nand autoencoders to construct 
 an efficient anomaly detection system. The model is proficient at learning
  the normal movement patterns within\nthe data\, effectively distinguishin
 g between normal and anomalous be-\nhaviour\, and exceeding 90% across all
  four evaluation metrics including\naccuracy\, precision\, recall\, and F1
 . However\, it may encounter chal-\nlenges in accurately detecting anomali
 es where they deviate slowly from\nthe expected movement patterns. Despite
  this limitation\, the model\ndemonstrates promising capabilities by pinpo
 inting the precise loca-\ntions of anomalous entries within the dataset. F
 urther investigation\,\nincluding refinement and optimization of the model
 ’s parameters and\ntraining process\, especially with memory-based NN-AE
 \, may enhance\nits ability to detect anomalies with greater accuracy and 
 reliability.\n\nhttps://events.chpc.ac.za/event/139/contributions/2419/
LOCATION:Boardwalk Convention Centre BICC.G-D2 - D2 Tsitsikamma
URL:https://events.chpc.ac.za/event/139/contributions/2419/
END:VEVENT
END:VCALENDAR